Although the MEG is complementary to the EEG, no method is yet available for optimally combining the MEG and EEG, for determining electrical sources in the brain. In the previous work a method of this type was introduced for determining evoked sources particularly, where MEG and EEG maps over the head are combined in four steps; initial tests were made and the results were good. In simulations using a single dipole source in a simple theoretical model of the head, the method could accurately locate the source if the noise was not excessive. In actual evoked responses, where only dipole-like maps were chosen (somatosensory N20, P25, P30, P55 and auditory N100), the method determined sources at reasonable locations; as a byproduct, the claim was supported that N20 and P30 are primarily tangential dipoles in the vicinity 3b. However, in those initial tests there were two main shortcomings: Not knowing the exact locations of the evoked sources (hence the accuracy of the method), and not testing the method on multi-dipole sources. It is here proposed to overcome these two shortcomings. First the method will be tested on a single dipole source in both a more realistic computer model of the head, and an experimental head model (saline-filled skull); because the source locations will be exactly known, the accuracy of the method will be seen and adjusted if necessary. As verification, it will be tested using actual MEG and EEG maps due to a single source of known locations; these will be of epileptic subjects (from another study) with known lesions. Then the method will be tested (and adjusted) on the visually-evoked P100, which is somewhat dipolar but the source can be "moved around" so that the method should follow it; because MEG sources should be in fissures, partial verification will be obtained by overlaying these sources on NMR images showing the fissures. Next, it will be tested and adjusted on multi-dipole sources in both the theoretical and experimental models. Finally, the method will be tested on evoked responses due to multiple sources; these will be components in the 30-50 msec range due to auditory click; again NMR will supply partial verification. These tests should open the way for use of the 4-step method. As a byproduct, they should again help resolve some problems concerning the evoked sources used here.